To an excellent approximation, a structure made entirely of string and weights has tension on the strings but no shear stress on the string. (This is because string is floppy — if you pull gently on a string, you’ll pull the other end toward you but not sideways.). If the whole structure dangles without moving, that means that all the forces (tension and gravity) balance everywhere. If you flip the sign of all forces, tension turns into compression, gravity pulls the other way, there is still no shear, and the whole thing still stays put. Now you have an upside down cathedral with upside down gravity, which is more or less the same thing as a right side up cathedral with right side up gravity.
There are plenty of ways this can go wrong. The little weights Gaudí used might not correctly model the weight of the stone. The whole thing might be unstable under inevitable sideways forces from wind and such, although one could blow on the string model to make sure it doesn’t start swinging to approximate this. And the string model probably doesn’t say much about the distribution of compression loads in the stone, especially if the overall structure is not statically determinate.
Taking this a bit farther, here are some simple examples illustrating stability and static determinacy.
Imagine a single weight dangling from two strings of roughly equal lengths attached near other on the ceiling. The weight will swing if you blow on it. If you turn the strings into (lightweight) compression elements and turn the whole thing upside down, it will be a bipod. The weight will indeed be balanced, but any slight perturbation will knock it over or will at least require shear forces to avoid this fate.
Now try this with three strings. The weight will resist swinging if you blow on it, and the cathedralized version is a stable tripod.
Now try four strings. If the string model is imperfect, one string might be slack. If you cathedralize it, you get a quadropod, and, if the construction is at all imperfect, the loads on the four legs might be wildly different and, if it’s too far off, the structure might wobble. This structure is statically indeterminate.
There is, strictly speaking, nothing strictly wrong with static indeterminacy, but it makes the analysis considerably more complicated.
Strings really do have some stiffness. In a piano, this causes harmonics to be slightly more than an octave apart, leading to stretched tuning. Cotton strings won’t be as still as metal strings, but the effect still exists.
Another issue is that Gaudí used a bunch of little weights instead of weighted strings. This means that he’s modeling stone columns with all the weight concentrated in a few places, which isn’t quite right.
There are plenty of ways this can go wrong. The little weights Gaudí used might not correctly model the weight of the stone. The whole thing might be unstable under inevitable sideways forces from wind and such, although one could blow on the string model to make sure it doesn’t start swinging to approximate this. And the string model probably doesn’t say much about the distribution of compression loads in the stone, especially if the overall structure is not statically determinate.